How to handle overflows in a register like DWT_CYCCNT

[LMI2]

Posted on September 06, 2016 at 20:40

There are examples counting clock pulses with the DWT_CYCCNT, but it will overflow in time.

Counting pulses is easy if there is no overflow, but with overflow I am sure there nice way of doing it. I could do something like this. if old value>new value Max count of DWT_CYCCNT-old value+ new value. But how to get the max in DWT_CYCCNT and what happens if the DWT_CYCCNT overflows twice or more.

[Tesla DeLorean]

Posted on September 06, 2016 at 21:09

Ask yourself what it is you want to measure, over what time intervals.

If you want to benchmark things under 24 seconds (@180MHz) then such durations and timeouts do not have worry about wrapping.If you want some long time base, use a slower counter, or a SysTick based counter counting off milli-seconds and seconds, for years.

[christoph2399]

Posted on September 06, 2016 at 21:31

never came up with the idea that i need to count cycles @ a 32bit controller. last time i debugged efficiency was on an 8bit.

for program flow 1ms timestamp is more than enough.

[waclawek.jan]

Posted on September 06, 2016 at 21:46

> never came up with the idea that i need to count cycles @ a 32bit controller. last time i debugged efficiency was on an 8bit.

>

> for program flow 1ms timestamp is more than enough. I envy your easy job. Oh, wait, no, I love real challenges. JW

[waclawek.jan]

Posted on September 06, 2016 at 21:46

> never came up with the idea that i need to count cycles @ a 32bit controller. last time i debugged efficiency was on an 8bit.

>

> for program flow 1ms timestamp is more than enough. I envy your easy job. Oh, wait, no, I love real challenges. JW

[LMI2]

Posted on September 06, 2016 at 22:07

A wonder, the forum did not crash.

I was planning to let counter to run continuously for some time and check the results via serial bus.